📖Introduction

Its hard to make predictions, especially about the future. -Yogi Berra

Reputation is "the beliefs or opinions that are generally held about someone or something." Reputation is an aggregate of direct reputation sources and secondary sources. Direct reputation sources include personal relationships and personally observable behaviors or evidence. Secondary reputation, which we will refer to as "reputation proxies," include credentials, past experience, and other forms of non-personally experienced measures of reputation or credibility.

Reputation is a foundational requirement for nearly all human interaction and the core of economic value exchange. Hiring an employee, trusting a news source, or picking a babysitter rely on reputation. Even Bitcoin and Ethereum rely on their reputations for trustlessness and security to promote adoption because most users are not personally auditing the technology and verifying this for themselves.

Reputation Proxies

Reliance on direct or indirect reputation sources is affected by community size and reputation type. Both of these influences are affected by the time window allotted for reputation measuring. As populations grow from small N groups, small tribes or communities, to large N groups, societies, the primary source of reputation increasingly shifts from direct sources to proxies. You select someone as a spouse based on your personal experiences, but you trust someone to provide your medical care based on whether they hold a degree from an established medical school.

As modern economies have shifted from primarily physical to knowledge-based, expertise has become more abstract, and difficult to quantify. You can directly observe someones capacity to build a chair or perform a trade, but it is much more difficult to objectively measure an individual's capacity for analyzing macro economic trends. This leads to a reliance on university degrees, employment history, or other reputation proxies to determine if someone is fit for a role. Population size and abstract expertise push towards reputation proxies on short time horizons, but long time horizons still trend towards primary reputation sources. When you have a few weeks to judge candidates you may hire based on a resume, but you promote or terminate someone based on their performance over a long period of time.

Problems with Digital Reputation Systems

Large communities and the trend towards knowledge-based economies have facilitated the rise of the use of digital systems as reputation proxies. Social media or other forms of online media are used by individuals to make public bets and build public reputations. Ryan Selkis' annual Crypto Theses are a perfect example of public bets used to build credibility. They are accessible to individuals of any background, provide a tie to identity, and time-bounds on predictions. However, using predictions as a basis for building reputation is highly susceptible to Sybil attacks and curation. It also often relies on already having an audience or needing to build an audience to make any reputation valuable. With the loose tie to immutable identity, it is trivial to create multiple accounts with contrary predictions and then only adopt accounts whose predictions were correct.

A similar effect can be used on one account by making predictions over a long time horizon and either deleting old incorrect predictions or only surfacing predictions that were correct. There is no built in, objective mechanism for measuring outcomes or prediction verification and it is trivial to manufacture fraudulent reputation. Public prediction statements are generally open, but can be exclusionary through censorship, deplatforming, legal jurisdictions, or non-quantifiable attacks on reputation (AKA canceling).

Problems with Prediction Crowd-sourcing

Though it is not their primary purpose, betting markets provide a fair system for producing quantifiable reputation and measure outcomes. They are uniquely useful in predicting future outcomes of discrete events through the wisdom of the crowd effect, take elections for example. They remove the "noise" of non-serious predictions by requiring financial stake to participate. The flaw in this system is that it is not tied to identity and difficult or unsuitable for curating reputation. This and other crowdsourcing solutions also fall short in their failure to factor in past performance into the predicted likelihood of outcomes. If betting odds are 99 yes to 1 no, but the single no voter has a past performance of 95% accuracy and the 99 yes voters are low 20% accuracy voters, then an outcome is most likely not actually 99% likely.